Alternative feature extraction approaches for car plate recognition;

Abstract

License plate detection is an important problem for surveillance systems and traffic applications. This work primarily focuses on feature extraction approaches for extracting candidate car plate regions. For this purpose, we start with restricting our feature extraction regions by use of a hierarchical edge detection method. Edge detection is based on discrete wavelet transform. Performances of Haar and Daubechies filters are compared on that basis. Three different methods are proposed for feature extraction: histogram feature, horizontal profiling on the edge map, and segment features followed by segmentation. The performance of each algorithm is presented by use of tests on ATS images. © 2005 IEEE.